A new intrusive method, combined of several independent objective metrics, has been developed for the evaluation of the quality of acoustic noise suppression in mobile communications. Extensive subjective data, including simulations of several noise suppression solutions in various noise environments, was gathered to serve as the benchmark for the metrics. Partial least-square regression and full cross-validation were used to establish the applicability of 26 metrics, that were making use of different measurement procedures, to predict the perceived quality. A Phase IV, vector-based preference model, was optimized to predict quality with a correlation of 0.95, resulting in an average prediction error of 8 %. Different measurement procedures appeared to contribute with a similar extent to the prediction ability of the optimized model.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.